Organizing Multipath Routing in Cloud Computing Environments

One of the objectives of organizing cloud systems is to ensure effective access to remote resources by optimizing traffic engineering (TE) procedures. This paper considers the traffic engineering problem in a cloud environment by using a multipath routing technique. The multipath routing algorithm is used to identify the maximum number of disjoint paths in the graph which overcomes the problem in the junction area estimation process. So, the algorithm forms a plurality of nonoverlapping and partially intersecting paths between any two nodes is proposed. Finally, the conditions for the formation of multipath virtual channels to ensure minimum build-time posts for the parallel transmission of its parts are also discussed. Keywords—Cloud Computing; Remote resource; Optimizing traffic engineering; Multipath routing; Disjoint paths; Parallel


INTRODUCTION
Traffic engineering (TE) is an essential tool for the provision of reliable, differentiated, and fast network services.According to the Internet Engineering Task Force (IETF), TE is roughly signified as dealing with the aspect of network engineering pertaining to problems of performance optimization and evaluation of Internet Protocol (IP) networks.Moreover, TE often deals with traffic demand by mapping different types of a given network topology to reflect changing network conditions by adaptively reconfiguring its processes.It is better than Quality-of-Service (QoS) routing in the sense that TE normally aims at highly efficient operational networks while meeting particular constraints, whereas the major aim of QoS routing is to meet particular constraints of QoS for traffic flow from a given source to a destination.
An essential part of a multipath routing framework is optimal routing.In optimal routing, source-to-destination traffic is split at tactical points to allow the gradual altering of traffic along alternative paths.The main aim of optimal routing is to avoid traffic, particularly along paths that are the shortest in terms of packet transmission time.For increasing input traffic, alternative paths are used to avoid an overload along the shortest path.So, the multipath routing algorithm is used to identify the maximum number of disjoint paths in the graph which overcomes the problem in the junction area estimation process.Thus, this paper proposes an algorithm to form a plurality of non-overlapping and partially intersecting paths between any two nodes.The conditions for the formation of multipath virtual channels to ensure minimum build-time posts for the parallel transmission of divided data parts like transmission, routing path are also discussed.This work considers the traffic engineering problem in a cloud-based environment by using the multipath routing technique.It is expected that multipath routing will improve the flow quality of streaming in cloud environments, without particularly considering the short flows with dynamic routing.In terms of resource control, multipath routing can direct strong traffic oscillations, route flapping and excessive signaling message overhead and so on, taking an account of topology changes due to the dynamic routing, despite its potential benefits and use in static routing.Outdated information routed by packets can direct to load oscillations; thus, the objective of TE can be attained by routing traffic demands along different types of multiple paths.
The rest of this paper is organized as follows: section II discusses the studies related to the research work, section III presents the proposed algorithm, finding the maximum number of disjoint paths, the protocol for finding the minimum of the junction area of the graph and the conditions for the formation of multipath virtual channels to ensure the minimum build-time posts for parallel transmission of its parts, section IV discusses and analyzes the simulation results and finally, section V concludes this work.

II. RELATED WORK
Cloud computing allows users to worry less about understanding the details of infrastructures, and focus on optimizing the appropriate services and resources in the computational complexity.At the same time, the application of parallel transportation management systems has become a popular topic of research in the field [1,2].
Cloud computing platforms that provide Infrastructure as a Service are a form of virtual machines (VMs) for users, and are based on shared infrastructure, hardware, and software.At present, modern network technologies of clusters, grids, and cloud computing [3,4] are widely used in virtual private networks (VPNs) [5][6][7], which are built, as a rule, on global computer networks.Virtualization is carried out at different levels: server, storage, and network.Virtualization on local networks forms a private cloud as a VPN with a star or tree topology [5][6][7][8][9][10].At the same time, fat-tree topologies or switch-centric networks are becoming critical components of data center networks (DCN).This topology is known as a nonblocking multi-path network that utilizes several equal-cost paths between adjacent layers to help eliminate bandwidth bottlenecks in the core layers, in addition to supporting largescale networks consisting of several thousand physical servers [11][12][13].www.ijacsa.thesai.orgIn a report, He et al. [14] emphasized the challenging task of network management as they grow in size and complexity.They reviewed several optimization techniques that have been applied to network management problems.By realizing that optimization problems in network management are induced by assumptions adopted in the protocol design, they argued that protocols should first be designed with optimization in mind, rather than optimizing existing protocols and principles by changing architectures.Maguluri et al. [15] used a stochastic model for load balancing and scheduling in cloud computing clusters.They assumed that jobs arrive at a cluster according to a stochastic process, and utilized virtual machines (VM) with a focus on resource allocation problems and scheduling VM configurations.They primarily contributed to the development of frame-based non-preemptive VM configuration policies, and claimed that these policies are nearly throughput optimal, in contrast to the widely used best-fit policy that is known to be throughput suboptimal.Their simulations indicated that long frame durations are throughput perspective by providing satisfactory delay performance.Recently, Manjur et al. [16] have proposed a unified storage allocation scheme (USAS) for VM.The proposed algorithm is able to allocate space dynamically according to the requests of users (e.g., OS images) and employs storage partitioning theory.
Similar to tele-traffic engineering methodology in heterogeneous networks (HetNets) proposed by Saied et al. [17], Chiesa et al. [18] considered the standard model of traffic engineering (TE) with equal-cost multipath (ECMP) and proved that "ECMP can provably achieve optimal traffic flow for the important category of CLOS datacenter networks" in contrast to the known approximation.They also addressed a shortcoming in ECMP in the suboptimal routing of large flows by presenting a suitable algorithm for scheduling with provable approximation, thereby shedding new light on the performance of TE with ECMP.
Similarly, Liu et al. [19] considered multipath routing specific to communication networks from a traffic engineering perspective in a multi-commodity setting through linear programming.They showed that a multipath measure (MPM) is zero or close to zero under certain traffic conditions and topological structures, hence implying that there is limited multipath gain compared to that in single-path routing.For the all-pair traffic case, multipath routing was observed to be advantageous for small networks.They claimed that the effective distribution of traffic in multipath routing is significantly better over network resources, which is believed to be somewhat in opposition "load sharing." In another report by Wang et al. [20], AMPLE, based on offline link weight optimization, was introduced.Using this, they were able to monitor network dynamics at short timescales, thereby coping almost optimally with unpredictable traffic dynamics.They also formulated a new proposal for achieving superior service quality and overall network performance in IP networks with reference to real network topologies and traffic traces.Gojmerac et al. [21,22] proposed another algorithm called Adaptive Multipath Routing (AMP) for dynamic traffic engineering on the Internet, with continuous load distribution within a network domain, hence offloading congested links in real time.They reviewed several methods and algorithms in this context, and presented important areas of application of AMP for emerging networking architectures.This finding was based on their earlier work, where AMP was used within autonomous systems.

III. PROPOSED ALGORITHM
A survey of the literature reveals that optimization theories need to be designed in order to develop a better organizing process suitable for multipath routing in cloud computing environments, for the analysis and design of various components of traffic management to realize an optimal and versatile traffic engineering protocol.Therefore, one of the main objectives of organization in cloud systems is to create effective access to remote resources by optimizing the procedures of TE for transmitting data via cloud computing.The construction of multipath traffic consists of the following tasks: 1) Formation of a plurality of paths with predetermined QoS parameters.
2) Organization of multipath virtual channels focusing on data transfer involving different types of traffic.
3) Management of the transfer of information.Therefore, this paper addresses the traffic engineering problem in cloud environments using the multipath routing technique for data transformation via a cloud structure.An algorithm is proposed to solve the problem of finding the maximum number of disjoint paths, and a protocol for finding the minimum of the junction area of the graph is presented.Finally, the conditions for the formation of multipath virtual channels to ensure minimum build-time posts for parallel transmission of its parts are also discussed.

A. Terms of paths adjacency
Data transformation in a cloud structure is carried out by using multipath routing, such as non-intersecting paths, and paths that have common nodes or links.Paths with common non-adjacent nodes are called intersecting paths, and those with common communication channels are called adjacent tracks.The choice of a set of paths depends on the required QoS information to be transmitted and the efficiency of the information transmission network.This involves considering ways of formulating adjacency conditions for an arbitrary graph G = (V, E).

Lemma 1.
Path P i =(V i ,E i ) and P j =(V j ,E j ) do not intersect under the following condition: where V i and V j are sets of vertices for paths P i =(V i ,E i ) and P j =(V j ,E j ), v 0i and v ei are the initial and final points of path P i =(V i ,E i ), and v 0j and v ej are the initial and final points of path P j =(V j ,E j ).Lemma 2. Paths P i =(V i ,E i ) and P j =(V j ,E j ) intersect when (2) Lemma 3. Paths P i =(V i ,E i ) and P j =(V j ,E j ) are adjacent when www.ijacsa.thesai.org The coefficient of intersection k ri , and path P i =(V i ,E i ) can be determine from the ratio of N r vertices in common with other ways to a variety of N Pi own vertices path P i (V i ,E i ), i.e., k ri =N r /N Pi .At k ri =0 leading to a condition that path does not intersect with other paths, while at N r =1, the path partially overlaps.
Accordingly, under the adjacency factor k ci , paths P i = (V i ,E i ) will be the ratio of N c common ribs to a plurality N Ei all the edges of the path, i.e., k ci = N c /N Ei .At k ci = 0, leading to a condition that the path is not adjacent, while at N c = 1, the path considered as a weak bound path.

B. Determining the minimal set of junctions
The maximum number of paths depends on network topology, the degree of the vertices of the network, and the set of values k ri and k ci .The maximum number of disjoint paths between two vertices v i and v j is determined by a graph of the minimum set of joint vertices V S = V/(V 1  V 2 ), i.e., the minimum set of vertices whose removal divides the graph G=(V,E) into two subgraphs: Determining the minimum set of junctions can significantly reduce the complexity involved in finding the set of disjoint paths for known combinatorial algorithms, such as Dijkstra's algorithm.In the formation of k paths, the complexity incurred is О(kN 2 ), where N is the number of nodes in the network.In this case, the paths between nodes , at first from the top v i to the vertices of set V S , and then from these vertices to vertex v j .The time complexity of the search for k disjoint paths in subgraph G 1 =(V 1 ,E 1 ), by using Dijkstra's algorithm, is О(kN 1 2 ), where N 1 is the set of subgraph vertices G 1 =(V 1 ,E 1 ).Accordingly, the time complexity of the search for . For example, when N= 90 and k=10, the complexity of the formation of 10 direct routes between two vertices is О(81000).As the graph divides G=(V,E) using a minimal set of junctions with N 1 = N 2 , the subgraph with 40 vertices and k=10 will have a complexity of О(2kN 1 2 ) = О(32000).In the latter case, the complexity is less by about 2.5 times than N 1 = N 2 condition.
Thus, the problem of finding the maximum number of nonoverlapping or partially overlapping paths can be reduced to the problem of finding a minimum set of junction graphs with the subsequent formation of disjoint paths to the heights of the minimum set of junctions.This reduces the complexity of the algorithm to form a plurality of disjoint paths.
The proposed algorithm determines the minimal set junction based on the procedure of forming a junction between two subgraphs where the number of N 1 vertices in subgraph G 1 =(V 1 ,E 1 ) varies from 1 to (N-1).This is a result of sequentially generating several sets V S , including the selected V Smin with minimum power h Smin .Forming a plurality of junctions in subgraph G 1 =(V 1 ,E 1 ) will help distinguish between internal and boundary vertices.

Vertices of set
) not adjacent to the vertices separating sets V s will be referred to as internal vertices.For a set of internal vertices Accordingly, edge e b k,j is the boundary edge.In the set of boundary vertices V b  V 1 , and the edges are belongs to the i vertex has the degree value is v i .
The number of tops of internal edges v k determines the internal degree S i k .In turn, an edge The number of external edge tops v k defines the outer degree S b k .The process of determining the minimum set of junctions V Smin involves the successive formation of set of vertices V S , and determining V Smin : From the vertices adjacent to the initial vertex v i , a set of vertices V S is formed, which in this case is V Smin .

On the basis of vertices v i  V 1 adjacent to the vertices of set
V o 1 , vertex set V S is formed.6.The power h S of the vertex set V S is calculated.

The power h S of set V S is compared with power h Smin of set
V Smin .If h Smin > h S , the set of junctions V S becomes V Smin.8.The graph G 2 = (V 2 ,E 2 ) is formed with a new set of vertices V 2 = V 2 /V S .

If V 2 ≠ {v i } return to Step 2. End
The process of determining the minimum set of junctions in the formation of a path between vertices v 0 and v 18 is shown in Fig. 1, and consists of the following steps: where , and the original graph G=(V,E).

C. Determining plurality of disjoint paths
It should be noted that the initial vertex between v i and the vertices of the set junction V Smin may contain several disjoint paths, the number of which is greater than or equal to the cardinality of V Smin .Between the vertices of junction set V Smin and final vertex v j , there may also be several disjoint paths.
In order to avoid operation directed enumeration characteristic of combinatorial algorithms of ways, a streaming algorithm to form paths from one node to multiple nodes on the basis of the "branch and bound" method is proposed.At the initial stage, the decision tree consists of primary vertices, e.g., vertices v 0 (see Fig. 1) and related vertices V b = {v 1 , v 2 , v 3 }, which in this case are boundary vertices of subgraph . In forming paths in a set V 0 , every time a vertex is added v i  V b , , having fewer external branches as compared to other boundary vertices.Accordingly, to a set V b , vertex v j is added adjacent to vertex v i , with minimal external degree S b j .Thus, a decision tree is constructed from the root in vertex v 0 until it has all disjoint paths to a given node.
Given this notation, the algorithm to form a plurality of paths from vertex v i to the vertex of a given set V z of vertices is as follows:

Form a set of boundary of vertices V b = {v j | j=1,2,…,k}, which
in this case is a set of vertices adjacent to vertex v i .

For j=1 to k, specify path P
, form the set of paths W 1 ={ P j }.V z adjacent to vertex v m , the formation of path P i to vertex v k concludes.9. Path P i is added to the set of paths.

If a set of external vertices V b ≠, go to Step 4. End
As an example, consider forming a plurality of paths between vertex v 0 and vertices v 7 and v 8 (Fig. 1), as follows:

End
The process of forming disjoint paths is shown in Fig. 2. The second step of the algorithm generates a path (Fig. 2a) between the initial vertex and adjacent vertices.The number of such paths is the degree of the initial vertex.In Step 9 of the algorithm, paths are formed (Fig. 2b) from the initial vertex v 7 to vertices v 7 , v 8 , and v 9 .The algorithm generates a plurality of disjoint paths (Fig. 2c) from the initial vertex to the ends of vertices.Thus, between vertex v 0 and vertices v 7 and v 8 are formed the following paths: A characteristic feature of this algorithm is that it forms a set of paths according to predetermined criteria for optimal QoS.In this case, the length L i of path P i -namely P 1 = P 2 =2; P 3 =3.The formation of the set of paths between the boundary nodes (v 7 ,v 8 ) and final vertex v 18 is carried out in a similar manner, starting with the final vertex:

Begin
Step
Between vertex v 7  V Smin and v 18 , there are paths P 7 ={v 7 ,v 13 ,v 16 ,v Thus, depending on the desired transmission quality, QoS parameters between vertices v 0 and v 18 may form the shortest path: for example, the path {P 5 , P 11 } of length L 5,11 = 4.The longest path is {P 1 , P 7 } with a length of L 1,7 = 7.In organizing, a parallel transmission path may be formed {P 5 , P 10 }, {P 2 , P 8 }, and {P 6 , P 11 }, of length 5.In this case, parallel to the transmitted part, the data will be collected without additional delay in the receiving node.

D. Determining the parallel transmission of paths
In general, between vertices v 0 and v 18 may be formed the following set of paths: The presence of a sufficiently large set of all possible paths makes easier the process of multipath transmission traffic.During the multipath transmission the QoS parameter has been maintained using the nature of the traffic requirements in the multipath virtual channel.
For example, if the data is divided into different pieces and those data has been transferred in to the parallel route for managing the data transfer delay like minutes IGRP and EIGRP, then the number of transmission is managed by RIP protocol.Thus, the difference in the path metric value for parallel transmission should be minimal.Figure 4 shows the assembly of three parts of the data transmitted by the same route metric as a case of delay of information transmission, where : T irepresents that the transmitted time of i -th data part, Ci-denotes that the treatment time (recording) i -th data part.In this case, the time required for data assembly (t 3 -t 1 ) is minimum and equal to 3(t 1 -t 0 ).At transmission delay of each data part on  t= (t 1 -t 0 ) relative to the previous part of the data (shown in Fig. 5), the time of whole data assembly remains minimal.In case of a delay  i > (t 1 -t 0 ) in transfer, the i-th part of the time required to assemble data parts is increased by t z = i -(t 1t 0 ), as shown in Fig. 6.The use of partially overlapping paths allows the formation of paths with similar metrics.For example, consider a set of paths Р i with metrics М i : Р 11 = {v 0 ,v 2 ,v 8 ,v 11 ,v 18 }, М 11 = 4; Р 12 = {v 0 ,v 3 ,v 6 ,v 8 ,v 10 ,v 14 ,v 18 }, М 12 = 6; Р 13 = {v 0 ,v 1 ,v 9 ,v 7 ,v 12 ,v 15 ,v 18 }, М 13 = 6.www.ijacsa.thesai.org The difference between the metrics is М 13 -М 12 = 0, М 3 -М 1 = 2, and М 2 -М 1 = 2, respectively.In this case, the time required to assemble the entire data (t 3 -t 1 ) (Fig. 7) is maximal, and is equal to 4  t.Thus, the possibility of the formation of various multipath virtual channels allows the optimization of the transfer of information in cloud computing.

A. End-to-end delay
The end to end delay is a measure which is used to calculate the average time taken for transmitting the packet in the network .It was calculated using different numbers of nodes, such as 50, 75, 125, 100, and 150.Each node setup incurred different simulation times, such as 100, 150, 200, 250, 300, and 350 (ms).The average end-to-end delay was as shown in Fig. 9.The proposed optimization of TE procedures showed promising results in terms of end-to-end delay due to a minimum delay for different kinds of nodes.11 shows that the total packet delivery ratio of the system, here the expected results are more or less same as the proposed system generated results which means that the proposed multipath virtual channels has ensured minimum build time posts for parallel transmission of its individual parts in cloud environment.www.ijacsa.thesai.orgCONCLUSIONS This paper proposed an algorithm to calculate the minimum junction area to determine the maximum number of disjoint paths and partially overlapping paths for transforming data via cloud computing.The proposed method of forming partially overlapping paths by creating disjoint paths to the heights of the junction allowed a significant reduction in complexity.The formation of multipath virtual channels based on QoS requirements made it easy to design and improve cloud computing traffic.The possibility of multipath virtual channel formation with the same transmission delay for each path ensures minimal assembly time of data due to the parallel transmission of its parts.A simulation yielded promising results in terms of end-to-end delay and packet delivery ratio.

6 .
j  V b , define vertex v m with the minimal external degree S b m .Move vertex v m to the set of internal vertices, v m  V 0 7. Form a subgraph G 1 (V 1 ,E 1 ) where V 0 = V 0  v m .8. If, among vertices v i , there is no vertex v k  V z adjacent to vertex v m , go to Step 9. If, among vertices v i , there is vertex v k  the formation of final path P 1 = {v 0 ,v 2 ,v 8 } */ Step 4: Forming new border set: {V 1 V 3 } /* a new set of boundary nodes */ Step 5: Paths /* formation of final path P 2 = {v 0 ,v 3 ,v 7 } .*/ Step 6: Forming new border set: {V 1 } /* a new set of boundary nodes */ Step 7: Selecting node V 1 (counter=1) /* selection boundary vertex with the minimum value of external degree */ Step 8: Selecting node V 9 (counter = 2) /* selection of external vertex with the minimum value of external degree */ Step 9: Paths /* forming a path from vertex v 0 to vertex v 9 */ Step 10: Forming new border set: {V 9 V 2 V 3 } /* a new set of boundary nodes */ Step 11: Paths /* formation of final path P 3 = {v 0 ,v 1 ,v 9 ,v 7 } .*/ Step 12: Border set: =  /* a new set of boundary nodes = */

Fig. 2 .
Fig. 2. Steps to form a plurality of disjoint paths

Fig. 4 .
Fig. 4. Assembling parts of data transmitted along routes with the same delay

Fig. 5 .
Fig. 5. Assembling parts of data sent along routes with almost identical delay

Fig. 6 .
Fig. 6.Assembling the parts of data sent along routes with long delays

Fig. 9 .
Fig. 9. Average end-to-end delay B. Packet delivery ratio Fig. 10 shows the percentage of packet delivery ratio with respect to increasing simulation time.It is clear from the results that packet delivery ratio increases as the time of packets produced by source increases.The average packet delivery ratio was calculated using different numbers of nodes, such as 50, 75, 125, 100, and 150.Each node setup required different simulation times, such as 100, 150, 200, 250, 300, and 350 (ms).The proposed optimization of TE procedures showed promising results in terms of packet delivery ratio.

Fig. 10 .
Fig. 10.Packet delivery ratio Fig.11shows that the total packet delivery ratio of the system, here the expected results are more or less same as the proposed system generated results which means that the proposed multipath virtual channels has ensured minimum build time posts for parallel transmission of its individual parts in cloud environment.